Hi @Yixin_Ma ,
In this context, you might be particularly interested in the upcoming MACT (Mesh-based ACT) framework developed by @chunhungyeh from our group / the MRtrix3 team: it’ll allow for some of the discussed alterations to be performed more directly, in a less hacky or even properly hack-free way. See his post below for some details on the topic:
Hi @schilkg1 ,
While it might be inappropriate to have a lengthy discussion within this post, I do wish to clarify some points respecting your 2nd and 3rd questions. Please see below.
What happens if the streamlines stop just short of labels (for example if my labels don’t exactly align or match the WM/GM boundary)? Will they not be counted in the connectome generation?
Good question. We did have spotted on the misalignment issue that you mentioned here; indeed, only ~25% of streamlines will be used in the connectome largely because of such misalignment. We describe it as a source data mismatch problem: the WM/GM boundary derived from tissue partial volume maps (based on tissue segmentation) used in tractography is not always consistent with the boundary of brain parcellation images used in connectome construction. Performing a 2-mm search from streamline endpoints (to find the nearest labelled voxel ) is the current MRtrix3 ’s default mechanism implemented in tck2connectome to deal with such inconsistencies (see the -assignment_radial_search in the help page of tck2connectome ). Using the radial search, ~85% of streamlines will contribute to the outcome connectome. If interested, please see this abstract for the relevant influences.
Does MRTrix have any capabilities for connectome generations using meshes as opposed to labels? For example mesh of WM/GM interface or mid-cortex mesh. Similarly, can meshes be used, for use with anatomical constrained tractography, or should this be converted to binary masks?
To avoid losing precision, my approach to deal with the source data mismatch is directly using high-resolution tissue surface meshes to unify the source data: the same labelled surfaces are used for both tractography and connectome generation, as described in your question. I called this method Mesh-based ACT (or MACT ), where streamline endpoints are the intersection points between streamlines and the labelled surfaces (cortical and sub-cortical WM/GM surfaces). When using MACT for connectome generation, there is no need to convert, for example, the vertex-wise labels of HCP-MMP into volume-based labelled images.
For more details, please see MACT abstract for reference. If you registered the ISMRM2017 meeting, I think my presentation video (0058 ) should still be accessible, there are more explanations on the background and the principle of MACT.
Finally, about using a mid-cortex mesh, I believe that it would only require a simple change in MACT. Coincidentally, I did once consider using a mid-cortex mesh as a default setting (rather than taking both inner and outer cortical GM surfaces). If you or anyone see any potential benefits or applications of using mid-cortex, I’d love to hear your insights and recommendations. So, even though MACT is not yet available in MRtrix3 , it has already and still received serious development by me and has been applied to a few studies. Any queries about MACT are very much welcome; please let me know via a personal message or an email.
Many thanks.
The original post (in its original context) can be found here: MRtrix tutorial available on OSF - #32 by chunhungyeh .
Cheers,
1 Like